Well, the 5500 isn't really an agp board if you dig into it. It's closer to a 66mhz pci video card with an agp connector

feipoa wrote on 2012-07-13, 00:08:

Anonymous Coward wrote:

VIA couldn't even get the basics right. I can only imagine how shitty one of their "server" chipsets must be.

That is my first thought as well, but I still have a glimmer of hope. Has anyone had ANY experience with the VIA Apollo Pro 266T chipset?

I don't have any dual tualatin boards, but I do have a Chaintech 6VJD Apollo Pro 266T and I'm very happy with it. I've never encountered any stability or compatibility issues, but my builds are very run-of-the-mill so keep that in mind. My setup has a 1.4GHz tualatin, 256MB of DDR400 running at 266MHz, an 80GB IDE Maxtor Diamondmax HDD, a Palit Geforce 3 Ti200, an Aureal Vortex 2 sound card and one 12MB Diamond Voodoo 2. It runs windows 98se, has loads of games installed, and so far it's run everything flawlessly. It also has a 3com PCI lan card but I's disabled, and I haven't had the PC on a network for over 2 years. I still pull it out from time to time when setting up a lan party, and it's served as my main win98 gaming PC from 2015 to 2016.

Performance-wise it's on par with my other Tualatin build which is based around an Abit ST6.

Boards with an Apollo Pro266(T) are realy rare. At that time Intel wanted to push the P4 and Sockel 423/478. So they gave the sockel 370 and the Apollo chipset with a Pentium III-S no attention. For my ASUS CUV266-DLS i can say, the AGP performance is normal.

MadYoshi wrote on 2023-01-19, 22:39:

Boards with an Apollo Pro266(T) are realy rare. At that time Intel wanted to push the P4 and Sockel 423/478. So they gave the so […]
Show full quote

Boards with an Apollo Pro266(T) are realy rare. At that time Intel wanted to push the P4 and Sockel 423/478. So they gave the sockel 370 and the Apollo chipset with a Pentium III-S no attention. For my ASUS CUV266-DLS i can say, the AGP performance is normal.

image108.jpg

3DMark2001SE_14239.JPG

wPrime_66.563.JPG

Now that is a nice, impressive, looking interior!

Seconded, very nice! Impressed!

MadYoshi wrote on 2023-01-19, 22:39:

Boards with an Apollo Pro266(T) are realy rare. At that time Intel wanted to push the P4 and Sockel 423/478. So they gave the so […]
Show full quote

Boards with an Apollo Pro266(T) are realy rare. At that time Intel wanted to push the P4 and Sockel 423/478. So they gave the sockel 370 and the Apollo chipset with a Pentium III-S no attention. For my ASUS CUV266-DLS i can say, the AGP performance is normal.

image108.jpg

3DMark2001SE_14239.JPG

wPrime_66.563.JPG

Holy hell dude! You beat the PIII-S 3DMark01 score I've been bragging about for years, and at stock CPU clocks to boot!
Yep, 266T is the only VIA P3 chipset with up to par AGP.

MadYoshi wrote on 2023-01-19, 22:39:

Boards with an Apollo Pro266(T) are realy rare. At that time Intel wanted to push the P4 and Sockel 423/478. So they gave the so […]
Show full quote

Boards with an Apollo Pro266(T) are realy rare. At that time Intel wanted to push the P4 and Sockel 423/478. So they gave the sockel 370 and the Apollo chipset with a Pentium III-S no attention. For my ASUS CUV266-DLS i can say, the AGP performance is normal.

image108.jpg

3DMark2001SE_14239.JPG

wPrime_66.563.JPG

Nice build! Mind telling me which CPU coolers you are using there? MSI Pro266TD's got a really awkward socket positions so i'm kinda limited at what I can use. I have tested it with my GF4 yesterday and it also behaved as expected.

As you can see, V5 goes perfect with it.

sirotkaslo wrote on 2023-01-20, 06:55:

Mind telling me which CPU coolers you are using there?

This are Cooler Master HHC-L61. I can also recommend Cooljag JAC 102C.

The lovely MSI Pro266TD. In my opinion, the Bios on it, is more fully developed.

Thanks, both are hard to get I see. I'll stick to OEM for now I guess.

I am playing around a Supermicro 370DLE.
Got two LinLin and a pair of SL5XL.
No AGP for me so I am using a PCIe to PCI adapter (I would love a PCIe to PCI-X one… Anyone??).
Using a GTS450 LP card for it has external power, so power delivery of pci slot is not an issue.
It works, but bandwith is pcie x1 and 3D Mark result is showing (6700 points)

Actually your GTS450 is running at a slightly slower speed: PCIe 1.0 x1 already can do 2,5 Gbit/s, while the Adapter probably could use maximum 66MHz/32 bit from the PCI-X bus, in which case it would be limited to 2,133 Gbit.
But this is still way better than my HD5450 PCI in my TYAN S2505t, as the PCI bus there runs at 33 MHz and the graphics card shares the bus with all other PCI components.
Anyhow, my system managed to get 6700 3DMarks in 2001 SE as well, so I assume that your adapter actually uses 32 bit as your graphics card should be way stronger than mine.
But I think that my system will be getting noticibly slowdowns if there is a lot of PCI activity in parallel (network card, harddisk access, etc).

Skorbin wrote on 2023-05-31, 12:40:

Actually your GTS450 is running at a slightly slower speed: PCIe 1.0 x1 already can do 2,5 Gbit/s, while the Adapter probably co […]
Show full quote

Actually your GTS450 is running at a slightly slower speed: PCIe 1.0 x1 already can do 2,5 Gbit/s, while the Adapter probably could use maximum 66MHz/32 bit from the PCI-X bus, in which case it would be limited to 2,133 Gbit.
But this is still way better than my HD5450 PCI in my TYAN S2505t, as the PCI bus there runs at 33 MHz and the graphics card shares the bus with all other PCI components.
Anyhow, my system managed to get 6700 3DMarks in 2001 SE as well, so I assume that your adapter actually uses 32 bit as your graphics card should be way stronger than mine.
But I think that my system will be getting noticibly slowdowns if there is a lot of PCI activity in parallel (network card, harddisk access, etc).

Nah… It was a bad bridge I was using.
Watch this:

A useless short analysis on PCIX to PCIe bridge performance on Tualatin

Here's my experience with MSI Pro266TD Master-LR (MS-9105) on a VIA VT865 chipset with DDR memory.

This is my first DUAL CPU motherboard designed with Tualatins in mind.

ATX Form Factor: 30.4cm x 25.4cm. Nice green PCB 😀

Specifications:
- 2 x 370 sockets with native Tualatin support (FC-PGA2)
- Apollo Pro266T chipset (VT8653+VT8233)
- 4 DDR memory slots!!! (supposedly suporting up to 4 GB of RAM)
- AGP 4x
- IDE RAID!
- LAN, PCI, USB and other standard crap
- No ISA 😜
More info here: https://theretroweb.com/motherboards/s/msi-pr … 266td-master-lr

My general thoughts:
- Runs fine with 1,4 GHz Tualatins(I bought it with two PIII-1133MHz tA1)
- It has a DDR support but it's very picky with RAM!
- BIOS should be more complex.
- Problematic with optimal configuration.
- Its performance is below my expectations.
- The placement of components on PCB is poor (hard to install big, solid radiators)
- It's nice that it works with just one CPU without unnecessary terminators.

BIOS Update:
Although I don't see any major differences between the available BIOSes, it's worth mentioning that the board has settings in the BIOS that block its update. You have to disable it in the BIOS in order to upload a new one.

Otherwise, you'll get an error:

Configuration, OC capabilities and performance:

A) Timing.
I like to push the hardware to the limit (sometime only to see how much it can go), but with this board, it's not that easy... All available BIOSes don't allow you to tighten some timings. The maximum you can set is 2-3-3-6 (CR2), which significantly limits the board's performance in games. Nevertheless, I figured out how to set timings to 2-2-2-5 (CR2) using WpcREDIT and this is what it looks like:

We have to edit offset 64 (set 00 01 10 10 = 2-2-2-5 CR2) where: (check the attached picture)

Bits 7 and 6 = 00 give x-x-2-5; 01 (x-x-2-6); 10 (x-x-3-5)
Bits 5 and 4 = 01 give 2-x-x-x; 00 (1.5-x-x-x supposedly...)
Bits 3 and 2 = 10 give x-2-x-x; 01 (x-3-x-x)
Bits 1 and 0 = better left unchanged, this is probably responsible for RAM Interleave

Unfortunately I couldn't figure out what is responsible for Command Rate and this is a big problem 🙁

B) Overclocking.
The advantage of this board is that you can adjust the FSB in 1MHZ increments from 133 to 166 MHz! Unfortunately, you can't raise the voltage for CPUs, so all we have left are Volt-MODs if necessary. I can assure you that the board reaches 166 MHz and is totally stable (even on CL-2-2-5). There is a high chance there is some more OC potential (similar platforms have OC options up to 200MHz), but it would require some modifications to go beyond 166 MHz options.

C) Performance (Windows XP SP3):
Performance is probably the biggest disappointment I've encountered while having fun with this board... For sure the makers of it had stability in mind in the first place, and even though the board has an AGP, I wouldn't expect a lot of frames in games. This platform has terribly low memory latency (107ns)!!!

This is what it looks like when you set the maximum timings in BIOS 2-3-3-6(CR2)

3DMARK 2001SE

A bit better with other memories (GEIL)

Note that these results are even worse than on an old VIA 693A chipset!
I expected at least 9500 points, but deep inside I was hoping for for some 9800 or even 10 thousand points (this is what you could get on SiS635T or Intel BX based boards) and here, even after overclocking the processor to 150 MHz, you don't get 10 thousand points!!

Anyway... there is still WPCREDIT where you can manually lower the timings 😉 Here are some improved results:

The timings are a bit better 102.4ns:

SiSoft Sandra:

3D Marks:

(as a reference take a look at those charts Project Tualatin 2018)

PERFORMANCE SUMMARY:

Cachemem Memory Latency - 102.4ns - 7th place (better than VIA 693A and worse than Intel 815)

SiSoft Sandra:
MEMORY CPU: 494 MB/s - 3rd/4th place (better than Intel BX - 468 and worse than VIA VT8633 - 526)
MEMORY FPU: 515 MB/s - 3rd/4th place (better than Intel BX - 473 and worse than VIA VT8633 - 528)

CPU DHRYSTONE: 4168 MIPS - 1st place (the best, then 3838)
FPU Whetstone: 2044 MFLOPS - 1st place the (best, then 2000)

PC MARK 2002:
4431 CPU - 2nd place (4438 - Intel BX and 4427 SiS 635T DDR)
2802 RAM - 2nd place (2839 - Intel BX and 2587 Intel I815)

3D Mark 2000 - 9685 points - 11th place (only Intel I820 was worse with 9371 points, still better than ALI M1651T with 10361 points)

3D Mark 2001SE - 9381 points - 8th place (better than ALI with DDR but worse than VIA 694T - 9481 points)

(CPU 1400 MHz WinXP, MAX settings NO SMP)
QUAKE III
1280x1024 = 184.8 FPS - 7th place (placed between 187.2 points on VIA 694T and intel I820; the first was SiS 635T with 197.2 FPS)
640x480 = 205.6 FPS - 8th place (placed between 205.9 points on IntelI820 and ALI DDR 194.6; the first was SiS 635T with 227.2 FPS)

SERIOUS SAM
OpenGL 1280x1024 = 59.3 FPS - 2nd place (after 57.4 FPS, the best was SiS 635T - 60.2)
Direct 3D 1280x1024 = 25.3 FPS - 12th place (the worst result, the next is 26.6 FPS - intel I820; the best was SiS 635T - 35.8 FPS)

D) Overclocking Performance:
You can expect a performance increase of around 20%

150MHz no problem:

166 MHz CL2-2-2-5(CR2)

GEFORCE 6800 GT:

I wanted to do a thorough comparison with the project TualGOD Project, however I don't have two SL6BY Tualatins with VOLTMODs...
The problem is that with Command Rate set to 2, the board slows down significantly. Such comparison wouldn't be truly fair and at the moment I have no idea how to lower it to CR1 🙁

To sum up:
- The board is definitely interesting (DDR1 sockets, 2x socket 370, there is IDE RAID!) and I think that every Tualatin fan should have at least one of those or a similar one!
- Bad performance caused by poorly developed BIOS and some problems with RAM support.
- Limited overclocking options.
- You can install up to 4 GB of RAM!
- IDE RAID support!
- Possible to reach 166 MHz FSB with CL2-2-2-6 timings - everything stable (no 1/5 divider though!)
- It will definitely perform much better in applications that utilize 2 CPUs 😉

Nice write-up. I have one of these boards in a case with the IDE RAID going well.

How do I VOLTMOD SL6BY CPUs? Does it only work with SL6BY CPUs? Thanks.

Pentium 3 with DDR? I didn't even know that was a thing. I have seen single professor boards with RDRAM though.

feipoa wrote on 2025-01-16, 13:02:

Nice write-up. I have one of these boards in a case with the IDE RAID going well.

How do I VOLTMOD SL6BY CPUs? Does it only work with SL6BY CPUs? Thanks.

Thanks! 😀 Have you also noticed how "slow" it is? 😁

This volt mod is very simple and works well with all Pentium III Tualatins. I haven't tried with Celerons though 😉
By connecting those two wires you gonna raise the voltage from 1,45 to 1,65.

Bear in mind that NOT ALL Tualatins need that much to run FAST!

Here 170 MHz FSB (1786 MHz CPU clock) is totally fine at the default 1,45Vcore 😀

Here you can see what is my current maxium result with SL6BY at 1,712Vcore: https://www.youtube.com/watch?v=a_t52EwdxTw - 1900 MHz with AIR COOLING!!!

Kahenraz wrote on 2025-01-16, 13:20:

Pentium 3 with DDR? I didn't even know that was a thing. I have seen single professor boards with RDRAM though.

Man... you should better check my previous posts here 😁 Including those two threads 😉
Best Tualatin Motherboard
ECS P6S5AT at 166 MHz

feipoa wrote on 2018-01-12, 21:37:

I'd really like to see a dual Tualatin which functions properly with the Matrox Parhelia PCI-X 256 card. You have this card to test out in your TRL-DLS?

Ok I reviewed this thread and noticed (in this thread) there was never a successful response here (in this thread) so I thought I would just repost this here.

I ran some test on my P3TDLE (dual 1.4) and my matrox parhelia 256 pci ("pci-x") with stock 3DMark01 to compare to your results: […]
Show full quote

I ran some test on my P3TDLE (dual 1.4) and my matrox parhelia 256 pci ("pci-x") with stock 3DMark01 to compare to your results:

OS: Win2k SP4 & DX9.0c

1. 64 bit / 66 Mhz = 6068
2. 64 bit / 33 Mhz = 5137
3. 32 bit / 33 Mhz = 4377 ( one of the regular pci slots and not one of the top 2 "pci-x" / pci64 slots)

While this gpu isnt anywhere close to as powerful as the ones you tested I thought you might be interested in the 66Mhz to 33Mhz results and threw in the 32bit result to compare to the 64bit result just because.

From this A useless short analysis on PCIX to PCIe bridge performance on Tualatin

I noticed how all the dual socket 370 boards with native Tualatin support are slow. What's the fastest dual Tualatin board?

Do you have a table which shows all the Tualatin pin bridge modifications and their corresponding running voltages?

GremFive: You found a Serverworks dual Tualatin that works with the Matrox Parhelia 256? that's great. Did you have to do anything special to get the drivers running properly?

I always thought the greatest serverworks board was the HE-SL chipset. the intel SDS-2 or asus TRL-DLS. I havent checked the asus board, but I know the supermicro/tyan/intel boards offer zero overclocking support.

There is a problem with the parahelia 256 on serverworks boards? what is this issue? I have a small pile of the PCI-X versions here for a secret project. ( i got them cheap before the ePay price went nucking futs)

It has been too long since I was testing the Parhelia 256, but I ran into problems on both variants of the ServerWorks chipset - some issue with the PCI bridge. The details are hopefully in my old thread here: Matrox Parhelia 256 PCI-X Performance